Semantic content clustering based on user interactions

Abstract

Various embodiments of an apparatus, methods, systems and computer program products described herein are directed to a Topic Engine. The Topic Engine captures a plurality of content identifier sequences. Each respective sequence represents an order at which a corresponding user account accessed content. The Topic Engine generates a plurality of clusters. Each cluster is associated with respective content identifiers appearing within a proximity to each other across the plurality of content identifier sequences of different user accounts. The Topic Engine obtains one or more sample content identifiers from at least one cluster via sampling the cluster. The Topic Engine extracts keywords from content represented by the one or more sampled content identifiers. The Topic Engine identifies a topic for the cluster based on the one or more extracted keywords.

What is claimed is: 
     
       1. A computer-implemented method, comprising:
 capturing a plurality of content identifier sequences, each respective sequence representing an order at which a corresponding user account accessed content, wherein at least a subset of the plurality of content identifier sequences includes respective content identifier sequences that include at least a first content identifier and a second content identifier, each respective content identifier sequence in the subset further associated with a different user account; 
 generating a plurality of clusters, each cluster associated with respective content identifiers appearing within a proximity to each other across the plurality of content identifier sequences of different user accounts, wherein generating the plurality of clusters further comprises: generating at least a first cluster associated at least with the first and the second content identifiers based on a recurring range of proximity of respective positions of the first and the second content identifiers appearing across the subset's content identifier sequences; 
 obtaining one or more sample content identifiers from at least one cluster via sampling the cluster; 
 extracting one or more keywords from content represented by the one or more sampled content identifiers; 
 identifying a topic for the cluster based on the one or more extracted keywords. 
 
     
     
       2. The computer-implemented method of  claim 1 , wherein generating a plurality of clusters comprises:
 generating a vector representation of each content identifier that appears in the plurality of content identifier sequences, the respective vector representation indicating a position in dimensional space; 
 generating a first set of clusters, each cluster in the first set comprising a density of content identifiers positioned according to corresponding vector representation; and 
 generating a second set of clusters, each cluster in the second set comprising a diffuse collection of one or more content identifiers positioned according to corresponding vector representation. 
 
     
     
       3. The computer-implemented method of  claim 2 , comprising:
 wherein generating the first and the second sets of clusters comprises: reducing a dimensionality of each vector representation prior to generating the first and the second sets of clusters; and 
 after generating the first and the second sets of clusters based on reduced dimensionality vector representations of the content identifiers, for each respective cluster:
 (i) obtaining an original dimensionality from a corresponding vector representation of each of the content identifiers associated with the respective cluster; and 
 (ii) reduce a dimensionality the obtained original dimensionalities of each of the content identifiers associated with the respective cluster; and 
 (iii) generate one or more subclusters of the content identifier based on the respective reduced original dimensionalities. 
 
 
     
     
       4. The computer-implemented method of  claim 3 , wherein extracting one or more keywords from content represented by the one or more sampled content identifiers comprises:
 sampling one or more contend identifiers from the one or more of the subclusters. 
 
     
     
       5. The computer-implemented method of  claim 4 , further comprising:
 obtaining an embedding representation for each of the one or more extracted keywords; and 
 generating one or more clusters based on the embedding representations. 
 
     
     
       6. The computer-implemented method of  claim 5 , further comprising:
 obtaining a centroid vector representation from one or more of the embedding representation clusters; and 
 determining one or more similar words associated with the centroid vector representation. 
 
     
     
       7. The computer-implemented method of  claim 6 , wherein identifying a topic for the cluster based on the one or more extracted keywords comprises:
 generating a ranked listing of the one or more similar words based on an extent of similarity of each similar word with respect to the centroid vector representation; 
 identifying the topic via applying zero shot classification to the ranked listing of the one or more similar words. 
 
     
     
       8. A system comprising one or more processors, and a non-transitory computer-readable medium including one or more sequences of instructions that, when executed by the one or more processors, cause the system to perform operations comprising:
 capturing a plurality of content identifier sequences, each respective sequence representing an order at which a corresponding user account accessed content, wherein a first content identifier corresponds to first content provided by a first user, a second content identifier corresponds to second content provided by a second user, the first and second content comprising differing formats; 
 generating a plurality of clusters, each cluster associated with respective content identifiers appearing within a proximity to each other across the plurality of content identifier sequences of different user accounts; 
 obtaining one or more sample content identifiers from at least one cluster via sampling the cluster; 
 extracting one or more keywords from content represented by the one or more sampled content identifiers; 
 identifying a topic for the cluster based on the one or more extracted keywords. 
 
     
     
       9. The system of  claim 8 , wherein generating a plurality of clusters comprises:
 generating a vector representation of each content identifier that appears in the plurality of content identifier sequences, the respective vector representation indicating a position in dimensional space; 
 generating a first set of clusters, each cluster in the first set comprising a density of content identifiers positioned according to corresponding vector representation; and 
 generating a second set of clusters, each cluster in the second set comprising a diffuse collection of one or more content identifiers positioned according to corresponding vector representation. 
 
     
     
       10. The system of  claim 9 , comprising:
 wherein generating the first and the second sets of clusters comprises: reducing a dimensionality of each vector representation prior to generating the first and the second sets of clusters; and 
 after generating the first and the second sets of clusters based on reduced dimensionality vector representations of the content identifiers, for each respective cluster:
 (i) obtaining an original dimensionality from a corresponding vector representation of each of the content identifiers associated with the respective cluster; and 
 (ii) reduce a dimensionality the obtained original dimensionalities of each of the content identifiers associated with the respective cluster; and 
 (iii) generate one or more subclusters of the content identifier based on the respective reduced original dimensionalities. 
 
 
     
     
       11. The system of  claim 10 , wherein extracting one or more keywords from content represented by the one or more sampled content identifiers comprises:
 sampling one or more contend identifiers from the one or more of the subclusters. 
 
     
     
       12. The system of  claim 11 , further comprising:
 obtaining an embedding representation for each of the one or more extracted keywords; and 
 generating one or more clusters based on the embedding representations. 
 
     
     
       13. The system of  claim 12 , further comprising:
 obtaining a centroid vector representation from one or more of the embedding representation clusters; and 
 determining one or more similar words associated with the centroid vector representation. 
 
     
     
       14. The system of  claim 13 , wherein identifying a topic for the cluster based on the one or more extracted keywords comprises:
 generating a ranked listing of the one or more similar words based on an extent of similarity of each similar word with respect to the centroid vector representation; 
 identifying the topic via applying zero shot classification to the ranked listing of the one or more similar words. 
 
     
     
       15. A computer program product comprising a non-transitory computer-readable medium having a computer-readable program code embodied therein to be executed by one or more processors, the program code including instructions to:
 capturing a plurality of content identifier sequences, each respective sequence representing an order at which a corresponding user account accessed content; 
 generating a plurality of clusters, each cluster associated with respective content identifiers appearing within a proximity to each other across the plurality of content identifier sequences of different user accounts, 
 wherein generating the plurality of clusters comprises:
 (i) generating a vector representation of each content identifier that appears in the plurality of content identifier sequences, the respective vector representation indicating a position in dimensional space; 
 (ii) generating a first set of clusters, each cluster in the first set comprising a density of content identifiers positioned according to corresponding vector representation; and 
 (iii) generating a second set of clusters, each cluster in the second set comprising a diffuse collection of one or more content identifiers positioned according to corresponding vector representation; 
 
 obtaining one or more sample content identifiers from at least one cluster via sampling the cluster; 
 extracting one or more keywords from content represented by the one or more sampled content identifiers; 
 identifying a topic for the cluster based on the one or more extracted keywords. 
 
     
     
       16. The computer program product of  claim 15 , comprising:
 wherein generating the first and the second sets of clusters comprises: reducing a dimensionality of each vector representation prior to generating the first and the second sets of clusters; and 
 after generating the first and the second sets of clusters based on reduced dimensionality vector representations of the content identifiers, for each respective cluster:
 (i) obtaining an original dimensionality from a corresponding vector representation of each of the content identifiers associated with the respective cluster; and 
 (ii) reduce a dimensionality the obtained original dimensionalities of each of the content identifiers associated with the respective cluster; and 
 (iii) generate one or more subclusters of the content identifier based on the respective reduced original dimensionalities. 
 
 
     
     
       17. The computer program product of  claim 16 , wherein extracting one or more keywords from content represented by the one or more sampled content identifiers comprises:
 sampling one or more contend identifiers from the one or more of the subclusters; 
 obtaining an embedding representation for each of the one or more extracted keywords; and 
 generating one or more clusters based on the embedding representations. 
 
     
     
       18. The computer program product of  claim 17 , further comprising:
 obtaining a centroid vector representation from one or more of the embedding representation clusters; and 
 determining one or more similar words associated with the centroid vector representation; 
 wherein identifying a topic for the cluster based on the one or more extracted keywords comprises:
 generating a ranked listing of the one or more similar words based on an extent of similarity of each similar word with respect to the centroid vector representation; 
 identifying the topic via applying zero shot classification to the ranked listing of the one or more similar words.